Synchrophasor Data Applications for Wide-Area Systems ∗

Luigi Vanfretti Joe H. ChowRoyal Institute of Technology (KTH) Rensselaer Polytechnic Institute (RPI)

Stockholm, Sweden Troy, New York, USAluigiv@kth.se chowj@rpi.edu

Abstract - The successful deployment of Smart Gridsat the transmission level will largely depend on the devel-opment and implementation of synchronized phasor mea-surement data-based applications and on the continuous im-provement of the information and communication infras-tructures supporting them. During the last decade severalphasor measurement unit (PMU) data applications, that willpotentially enable Smart Transmission Grids (STGs), havebeen proposed. In this paper we provide an overview on twoof these applications, devoted to PMU-only state estimation,and modal analysis, as well as work related to developingsome supporting IT infrastructure. Rather than discussingdetails of applications, this paper aims to give highlights ofthe relevant features of these applications in the context ofSTGs, and outlines additional research efforts necessary forphasing in these tools into practical use.

Keywords - syncrhonized phasor measurements, wide-area measurement systems, wide-area control systems,wide-area protection, PMU-data applications

List of Acronyms

PMU Phasor Measurement UnitPDC Phasor Data ConcentratorSTG Smart Transmission GridWAMS Wide-Area Monitoring SystemsWACS Wide-Area Control SystemsWAPS Wide-Area Protection SystemsWAMPAC Wide-Area Monitoring, Protection

and Control SystemsPSE Phasor State EstimatorICT Information and Communications

TechnologiesQoS Quality of ServiceSPDC Super-Phasor Data ConcentratorMPLS Multi-Protocol Label SwitchingICCP Inter-Control Center Communication

ProtocolISO Independent System OperatorTSO Transmission System OperatorWECC Western Electricity Coordinating CouncilTVA Tennesse Valley AuthorityBPA Bonneville Power AdministrationFIPS Flexible Integrated Phasor System

1 Introduction

THE successful transition of today’s bulk power de-livery system into “Smart Transmission Grids” (i.e.

“Smart Grids” at the transmission level) will largely de-pend on the development and implementation of synchro-nized phasor measurement data-based applications, andon the continuous improvement of the information andcommunication infrastructures supporting them [6].

Over the last eight years many wide-area monitoringsystems (WAMS), have been deployed across the world[19] and are being continuously upgraded to meet un-precedented performance [8], delivery [5] and storage [28]requirements. At the same time, the number of availableinstallations of phasor measurement units (PMUs) contin-ues to increase [19], and several PMU-data applications,that could potentially enable Smart Transmission Grids(STGs), have been proposed [9]. In this paper we providean overview of two of these applications, synchronizedphasor measurement-based state estimation [24], and elec-tromechanical mode estimation [25], and on recent pro-posals for phasor data concentration [3].

Wide-area control systems (WACS) and wide-areaprotections systems (WAPS) and their applications are im-portant for the development of STGs. While we brieflydiscuss the elements of WAMS and WACS, this paper isdevoted as a review of a few wide-area monitoring el-ements, applications and related infrastructure. The re-mainder of this paper is organized as follows, Section 2discusses how WAMS can help achieve Smart Grids at thetransmission level, the architecture used for WAMS, anda phasor data concentrator design that can be used withinWAMS systems and as a Phasor Gateway. Next, in Sec-tion 3 different approaches for state estimation with PMUsare discussed. In particular, it is shown how the PhasorState Estimation (PSE) approach can be used to deal witha specific type of bad data observed from real PMU mea-surements from the US Eastern Interconnection. A secondimportant application of synchrophasor data is discussedin Section 4, and emerging challenges for mode estima-tion using mode meters are outlined. Finally, a conclusionof this review paper is given in Section 5.

∗Corresponding author: Dr. Luigi Vanfretti.L. Vanfretti is with the Electric Power Systems Division, School of Electrical Engineering, KTH Royal Institute of Technology, Teknikringen 33, SE-10044 Stockholm, Sweden.Joe H. Chow is with the Electrical, Computer & Systems Engineering Department, Rensselaer Polytechnic Institute, Troy, NY 12180, USA.

Data Alignmentand

Concentration

Monitoring

AdvancedDisplays

Near Real-TimeDynamic Security

Assessment

Early WarningSystem

AutomaticDetermination ofControl Actions

Decision/Control Support System

Smarter OperatorDecision Support

Smarter OperatorControl Actions

Transmission System Control Center

Com

municationNetworks

Sync. Timing

Controllable and Protective DeviceMeasurement and Status Data Sets

Synchonized Phasor Measurements

Smart-Automatic Control Actions

Data Storage

PDCPMU

Figure 1: A centralized model of a smart transmission grid

2 Wide-Area Systems: The Enabling Technologiesfor Smart Transmission Grids

2.1 Synchrophasor-Enabled STGs

A STG should be considered as more than a hybridsystem of power, measurement, and communication ap-paratus liked through an information and communicationtechnologies (ICT) system handling a vast amount of data.At a minimum, a STG should make use of this data in or-der to exploit all the available “observability” and “con-trollability” in a power system through closed-loop feed-back control, and to coordinate system control with pro-tection. As such, a STG can behave as a “self-healing”system, or at least provide mechanisms so that the powergrid can be operated more securely through increasedawareness.

To accomplish these challenges, STGs should containthe elements outlined in Fig. 1, where a conceptual dia-gram of a “centralized model” for a Smart TransmissionGrid is shown. In such STG synchronized measurementsare obtained at transmission substations through not onlyfrom PMUs, but also from other highly accurate time-synchronized measurement systems retrieving data fromcontrollable devices and protective device “informationsets” (i.e., all available information from within a protec-tive relay).

This plethora of data is sent through communicationnetworks, received and concentrated at a Decision andControl Support System that determines appropriate pre-ventive, corrective and protective measures. This supportsystem is the cornerstone for enabling STGs using syn-chrophasor data. It is here where the newly developedanalysis techniques will produce “smarter” operator deci-sions allowing the power system to operate more securely,efficiently, and reliably. The decisions determined by thissupport system will then support operators at control cen-ters to take “smarter operator control actions” or even de-vice “smart-automatic control/protective actions”. Theseactions are translated into feedback signals that are sent

through communication networks to exploit the controlla-bility and protection resources of the power system.

The centralized model is conceptually very similar towide-area systems currently being developed, however,automatic feedback control is limited in today’s WAMSsystems.

Note that although the diagram shown in Fig. 1 is acentralized model, there can be other more decentralizedmodels for STGs. A “decentralized model” is also possi-ble, and most likely more appropriate to deal with the largeamount of data involved. While this is out of the scope ofthis paper, it is important to realize that in such “decentral-ized model”, data delivery can be accomplished through apublisher-subscriber model, such as GridStat [11] whichuses Quality of Service (QoS) concepts for these purposes.Feasible approaches considered in [6, 11, 5] have great po-tential and should be further investigated.

It must be realized that regardless of the chosenparadigm (centralized or decentralized): the ICT systemwill be determined by the requirements from different ap-plications using PMU data, which in turn need the ICT in-frastructure to be developed and tested. The fact that bothICT aspects and specific requirements of each PMU dataapplication need to be considered simultaneously if reli-able applications are to be deployed deepens this dilemma.Hence, this dilemma also outlines the most important re-search challenge for wide-area systems. A possible solu-tion to find the appropriate ICT architecture that fits theneeds of future PMU applications is to develop a research,development, and testing platform where this issues canbe addressed. Such a platform would have a tremendousvalue for developing PMU-based applications supportingthe vision of Smart Transmission Grids explained above.

2.2 Wide-Area Monitoring, and Control Systems Ele-ments

Wide-area systems comprise a set of measurement de-vices, communication devices and links, and computersystems.

PMU PMU PMU

PDC

PMU PMU PMU

PDC

SPDCApplications Storage

Applications

Storage

Applications

Storage

LocalApplications

Local Storage

Local

CentralCentral

External

PDC

Figure 2: Generic Architecture of Wide Area Monitoring and Control Systems

Figure 2 presents a generic architecture which is com-mon today PMUs are measurement devices providingtime-synchronized phasor data; phasor data concentrators(PDCs) and super-data concentrators (SPDCs) are com-puter systems receiving, processing, and sending out re-sults to applications or storage, or signals for control; fi-nally communication is typically accomplished throughMulti-Protocol Label Switching (MPLS), virtual privatenetworks [21], or Frame relay circuits [7].

As indicated in Fig. 2, PMU data can be stored and ex-ploited at a local, central, or external layer. The local layeraddresses sub-station level applications and data handling.A PDC may be used, but it is not strictly necessary. At thecentral layer, different communication mechanisms maybe used, and thus the data from different locations may notarrive simultaneously. Hence, the most important func-tions of a PDC are to gather data from different locations,provide time-aligned data for applications, and create auseful archive of all the available measurements. In ad-dition, it is very common to find that at this central layersome PDCs have been supplemented with other functionssuch as historian capabilities, monitoring and control ap-plications. PDCs from different utilities or ISOs may alsoshare data among them to support interconnection-wideapplications. This is the external layer, that is used inlarge scale power systems as a further hierarchical levelfor wide-area applications and monitoring. To this ex-tent a Super Phasor Data Concentrators (SPDC) can beused to perform similar functions as PDCs, and to collectmeasurements over an entire network. Bidirectional com-munication is needed for data requests, configuration, andother interoperability functions.

This particular area of phasor measurement technol-ogy is continually evolving and will be subject to manyimprovements in the near future. In North America,NASPINet is being developed. The data delivery systemcomprises a publisher-subscriber model where publisheddata (by PMUs or PDCs) is given different QoS guaran-tees depending on the subscriber (phasor data application)[5].

2.3 Phasor Data Concentrators and Phasor Data Sys-tems

From the previous discussion it is evident that PDCsplay a crucial role in WAMS. Although phasor concen-

trators have been available in the Western Electricity Co-ordinating Council (WECC) power system [13] since themid-90’s, it’s not until recently that they have becomemore available [19]. Earlier PDCs, such as the BPA PDC,were specialized standalone units [16]. However, in Oct.2009, the Tennessee Valley Authority made openly avail-able the source code of the TVA SPDC under the nameopenPDC [28], and it is expected that it will acceleratethe use of synchrophasors in more utilities. Nevertheless,it should be acknowledged that most transmission oper-ators do not currently have the ability to process, store,and utilize the data from their own PMUs, and dependedon concentrators located elsewhere including those usingopenPDC.

FIPS

A Flexible Integrated Phasor System (FIPS) developedto provide an accommodating and lower cost alternativefor system operators and transmission owners to accesstheir phasor data was first presented in [3]. An overviewof FIPS is shown in Fig. 3. In the design stage of FIPS,several important issues came into consideration. Cur-rently, PMU data is routed to large data concentrators incentral locations. This data may then be routed back to in-dependent system operators (ISOs) and transmission own-ers (TOs) via existing networks such as those based on theInter-Control Center Communication Protocol (ICCP) [1].Systems configured in this way have considerable commu-nication overhead, and data storage at the central data con-centrator is a difficult problem, especially as the numberof PMUs increases [28].

Moreover, in existing synchrophasor systems, networkissues (such as latency) can cause data loss. Reliable trans-port protocols, which detect and attempt to correct trans-mission errors, are adopted to combat this problem.

Efficient storage of synchrophasor data is also impor-tant. Existing databases may require large indexes or con-siderable time to extract a subset of the data. A new datastorage scheme based on simple, flat files eliminates theseproblems.

Hence, FIPS is designed as a phasor system capablereceiving, sharing, and storing data efficiently, and servesas a foundation upon which synchrophasor applicationscan be built. FIPS contains communication components,

a database, and a data alignment engine, permitting it tofunction as a phasor data concentrator. Also, FIPS con-tains interfaces to enable the development of applicationsusing the stored and real-time data.

Synchrophasors and their applications represent a cru-cial part of a future STG, and FIPS was also built keepingin mind the objective of facilitating research on PMU-dataapplications. Hence the availability of systems such asFIPS will enable these applications to be rapidly devel-oped. Further details on the features of FIPS can be foundin [3].

Protocol

Con

version

Protocol

Con

version

Protocol

Con

version

ReorderingBuffer

PMU

PMU

PMU

TimeAlignment

Protocol

Con

version

Protocol

Con

version

GPSSatellite

Authentication

Authorization

MetadataStorage

DataAccessGUI

MatlabAPI

Manag

ementInterface

Tim

eSeriesDatabase

FIP

S

Real-Tim

eApp.

Real-Tim

eApp.

DataAccess

Off-lineApp.

Admin.PC

Figure 3: An overview of FIPS

Phasor Gateways

NASPInet [14] represents a peer-to-peer architecturefor the distribution of phasor data and consists of an entitycalled a data bus to which phasor gateways are connected.These phasor gateways communicate with each other toexchange data on an as-needed basis.

FIPS can provide the functionality of a phasor gatewaywith the addition of a few network services. First, a ser-vice must be provided to allow access to metadata, suchas channel identifiers. Remote access to the database mustbe available, and a means to stream real-time data as it isreceived is also necessary. A system for authentication ofremote systems is necessary. But these functionalities arestraightforward to implement given a robust foundation.

An example network topology for peer-to-peer datasharing is shown in Figure 4. The FIPS server at the topof the diagram might be located at a large regional controlcenter such as an ISO. The servers at the bottom might be

located at transmission owners. All data is routed throughthe MPLS links to the central location, and the router atthat point may transfer data between the two TOs, if thatis desired.

When operating as a phasor gateway, the possibility ofmulticast for data transport should be considered. These,and other aspects, are studied in [3].

FIPS

FIPS FIPSMPLS

Physical circuitsMPLS routed paths

Figure 4: MPLS network topology for simple peer-to-peer data sharing

3 State Estimation using Phasor Measurements

3.1 Conventional State Estimation including PMU data

There has been much discussion about using phasormeasurement data for state estimation. Most of the effortshave been directed into incorporating PMU measurementsinto conventional state estimators by merging it with con-ventional ICCP data [27, 12, 15]. While combining thesedifferent data types augments the measurement set of theestimator [4], it has the drawback of losing most of the in-herent characteristics of the power system dynamic naturecaptured by PMUs.

Although some improvements of including PMU datainto conventional SEs have been identified [10], the ap-proach to implement PMU data into conventional stateestimators needs to be revised to harmonize ICCP (staticdata) with PMU data (dynamic data): it is necessary toaddress how to select the correct time window from PMUdata that encompasses time-skewed data from ICCP, andhow to filter the PMU data to better match the static be-havior of conventional measurements.

There have been several proposals for exploiting pha-sor data such as hierarchical state estimation [12, 26], anddistributed state estimation [15], these approaches takeinto account detailed level of granularity of the the powersystem and combine both traditional RTU data with PMUdata. While improving the performance of conventionalstate estimators is important, a wide-area state estimatorcapable of capturing the power system dynamics embed-ded in PMU data would be an attractive solution for mon-itoring of large interconnected networks. In the reminderof this section, methods for PMU-only state estimationwhich focus on wide-area visibility are outlined.

3.2 Approaches for PMU-only State Estimation

There are basically two approaches for state estima-tion using only PMUs, the commonly known as linear

state estimation approach introduced in [20], and the pha-sor state estimation approach presented in [24]. Differentfrom conventional state estimators, these approaches donot use power flow models in their algorithms. Instead,being more natural to phasor measurements, Kirchoff lawsare used.

Moreover, the aim of these state estimators is to per-form state estimation for each snapshot of the system, i.e.,30-60 times every second, and to provide an independentstate estimator instead of replacing conventional ones al-ready in use. Nevertheless, the solution approaches dif-fer in that the linear state estimator uses a weighted linearleast-squares estimation solution, while the phasor stateestimator uses a non-linear WLS estimator to deal withinherent PMU data errors in the phasor angles. Both meth-ods are briefly explained and contrasted below.

3.3 Linear State Estimation

It was first realized in ( [22] - see Conclusions) that itis feasible to construct a linear state estimator by directlymeasuring voltage and line current phasors using PMUs,and by reformulating the WLS problem in terms of com-plex voltage and currents. In a later publication [20], itis described how to develop this purely PMU- based esti-mator, which is sometimes refer to in the power systemscommunity as a “state solver”.

While we do not present this method in detail, it shouldbe noted that in extremis direct state measurement can beaccomplished by deploying PMUs at all system nodes, inthis case the measured current phasors would provide re-dundancy in the measurement vector. Regardless of re-dundancy, this approach does not take into account the de-terministic behavior of certain PMU-data errors [24]. Oneof such possible errors is shown in Fig. 5, which presentsPMU data from the US Eastern Interconnection† from2008. Here PMU 1 presents angle shifts or “jumps” thatare not present in PMU 2 or PMU 3. These angle jumpswill have an important impact on the solution from the lin-ear state estimator. Moreover, the speculation that directstate measurements can be achieved by PMUs comes intoquestion with the existence of these data errors: a state es-timator capable of inherent bad-data correction is needed.

04:50:00 04:55:00−10

0

10

20

30

40

50

60

May 8, 2008 − Time (HH:MM:SS)

Volt

age

Angle

(deg

)

Voltage Angle Errors in PMU Data

PMU 1

PMU 2

PMU 3

Figure 5: Typical Angle Error in the Measurement Voltage Phasor fromPMUs (PMU 1 shows angle errors, while PMU 2 and PMU 3 do notshow angle errors)

We illustrate this drawback with the example powersystem in Fig. 6 in which all voltage and current phasorsare monitored by PMUs. First, PMU data is emulated asfollows: a dynamic response is computed from the powersystem model, errors and noise are considered by addingwhite Gaussian noise to the dynamic response, and finally,a snapshot of all phasors is taken and an angle jump of 7.5◦

is added to the voltage phasor angle. The linear state esti-mation solution is obtained from the procedure and designof measurement weights described in [20].

#2

G1

L1

#1

#3#4

L2

I12m I21m

I24m

I23m

I42m

I43m

V1m

V2m V4m

Phasor Measurement Unit

Current Phasor Measured by the PMU

V3m

I34mI32m

Figure 6: Example power system with all voltage and current phasorsmonitored by PMUs

# θtrue θmeas θse TrueResidual

Meas.Residual

1 20 19.99 19.97 0.03 0.0242 17.51 17.51 17.49 0.02 0.0213 -1.46 -1.46 -1.46 0.00 0.001

4 -1.46 6.03 6.02 7.49 0.009Table 1: Effect of phasor angle “jumps” in the voltage phasor estimatesfrom linear state estimation

Table 1 shows the results from this simple experiment.The superscript “true” correspond to the uncorrupted sim-ulation, the “meas” (measured) corresponds to emulatedPMU data, and “se” corresponds to the linear state esti-mation solution. Perhaps the most relevant issue to pointout is the difference between the true (fourth column) andmeasurement residuals (fifth column). Observe that whilethe true residual detects the presence of the angle jump ashighlighted by the last row of Table 1, the last measure-ment residual does not. Therefore, in this example the lin-ear state estimator approach has given an invalid solution.Moreover, this issue makes traditional bad-data detectiontechniques that depend on the measurement residual [17]inadequate to deal with this type of bad data. More impor-tant, this kind of bad data has a deterministic behaviour(although it occurs randomly), and knowledge of this char-acteristic can be exploited as explained below.

3.4 Phasor State Estimation

This state estimator approach relies solely on PMUdata [24], and has been designed to serve as an estimatorof unmonitored buse which can perform phasor angle biascorrection given measurement redundancy. From analy-sis of real field PMU data [24], we have observed several

†The names of the locations are not shown for confidentiality.

PMU data errors such as those in Fig. 5, which may bedue to a variety of reasons [24].

Phasor Estimated

Bus Voltage

Phasor Estimated Bus with Redundancy

Online PMU

Current phasor from

online PMU

1 Bus No.

1

2

3

4

5

6

7

(2) Voltage Phasors

Figure 7: A portion of American Electric Power’s (AEP) TransmissionSystem with PMU Observability and Redundancy for Phasor Angle Cor-rection

Phasor angles are typically computed using some sig-nal processing techniques over one cycle or portions of acycle to minimize the impact of quantization. The phase isalso affected by the length of potential and current trans-former cables. External synchronization issues with re-spect to GPS receivers and internal synchronization issuesdue to computational burden may also induce time delays,which translate into a phase lag in the measured data. Wehave observed random phase jumps of multiples of 7.5◦‡

and random saw-tooth behavior. Equally important, wehave also observed that if such a phase bias occurs in aparticular channel of a PMU, then the same bias will ap-pear in all the phase channels.§ Thus the polar coordi-nates provide a preferred setting for using weighted least-squares (WLS) techniques to correct the biases in the mea-sured phasor data. The phase bias is a form of bad data,which is not part of conventional SCADA data. Here themagnitude is assumed to be correct, although it is still sub-ject to normal calibration accuracy.

0 10 20 30 40 50 60

1

2

3

Time (sec)

Vo

ltag

e A

ng

le (

deg

) VOLTAGE ANGLE

Θ4m

(Meas.)

Θ4

e (Est.)

0 10 20 30 40 50 60

−1

0

1

2

Time (sec)

An

gle

Sh

ift

(deg

)

ANGLE SHIFT

Φ4 (Meas.)

Figure 8: Bus 4 measured and estimated voltage angle using the phasorstate estimator approach

The details on this phasor state estimator approach aregiven in [24]. However, it is useful to illustrate the ca-pability of this estimator to detect and correct PMU dataerrors as the ones described above. We have used a dataset that illustrates the capability for angle correction. Theresulting estimates provided by the PSE are shown in Fig.8 for Bus 4, and the measurement residuals for the voltageangles, eθi = θim − θi, are shown in Fig. 9. Angle biasesare significant in some channels, possibly due to firmware

and software issues in obtaining the phase of a measuredquantity as shown in the voltage angle residuals in Fig. 9,the angle jumps are effectively detected by the angle shiftvariable of the method as shown in Fig. 8. This is relevant,as these measurement residuals can be effectively used forbad data detection [17].

Observe from Fig. 8 that the voltage and current pha-sors measured at Bus 4 present an angle error. The voltageangle measurement shows an undesirable saw-tooth be-havior – a ramp with a periodic reset, which was probablydue to the internal clock synchronization with the GPS sig-nal. The PSE approach is able to remove this undesirableerror and to correct the data. We should stress that withoutthe angle bias correction capabilities of the PSE methodthe estimation process will converge to invalid solutions.In addition, necessary redundancy conditions must be metas described in [24].

0 10 20 30 40 50 60−2

−1.5

−1

−0.5

0

Time (sec)

Vo

ltag

e A

ng

le R

esid

ual

s (d

eg)

θ2m 1

− θ2θ2m 2− θ2θ5m 1− θ5θ5m 2− θ5θ6m− θ6

0 10 20 30 40 50 600

5

10

15

Time (sec)

Vo

ltag

e A

ng

le R

esid

ual

s (d

eg)

θ3m− θ3θ4m− θ4θ7m− θ7

Figure 9: Voltage phasor angle measurement residuals

4 Electromechanical Mode Estimation fromAmbient Data

The application of advanced signal processing tech-niques to power system measurement data for the estima-tion of dynamic power system properties has been a re-search subject for over two decade [13, 2]. These tech-niques have been proven successful for the estimation ofmode properties, i.e., mode frequency, damping, and morerecently, mode shape. Estimates of mode frequencies anddamping are indicators of power system stress. A declin-ing value of these indicators point to a detriment on thecapacity of a power system to operate reliably, and suchcondition may lead to a system breakup [13]. Mode shapeestimates can give a direct indication of the major areas ofa power system contributing to the oscillations in a spe-cific mode frequency, and therefore may be used in the fu-ture for determining control actions [2]. There are severaltypes of signal processing techniques available to providethese estimates: some are appropriate for transient signals,others are adequate for ambient signals, and still others

‡A 60 Hz signal measured at 48 points per cycle will result in an error of 360◦/48 = 7.5◦ if a data point is skipped.§All phase channels use the same GPS clock signal and identical digital signal processing code.

are used when a known probing source is used to excitethe power network [13]. Some of these techniques havebeen implemented in off-line analysis software, and arenow being incorporated into software tools used in controlrooms for monitoring the near real-time behavior of powersystem dynamics [18].

f2 = 0.18

f3 = 0.28

f4 = 0.37

f5 = 0.46

f6 = 0.55

f8 = 0.74

f9 = 0.83

f13 = 1.20

f12 = 1.11

f16 = 1.48

f17 = 1.57

f18 = 1.66

Figure 10: Welch Spectrogram for the P1 signal from Radsted in Den-mark. The red colors represent maximum values and the blue colorsrepresent minimum values of the power spectrum density [dB]. The timeis given in hours in (UTC) starting from 00:00:00 hrs, local time is givenin UTC+1 hr. on 03-20-2008.

Nonparametric and parametric power spectrum esti-mation techniques can be used to estimate the power spec-trum from synchrophasor measurements (PMU data). Theautomated and continuous application of this identifica-tion process is known as a “mode meter”. As a resultof applying these techniques and by focusing the analysisin the range of 0.1 to 1 Hz, the electromechanical modesare manifested as visible peaks in power spectrum den-sity estimate (PSD). Narrow peaks in the estimated spec-trum indicate light damping, and broader peaks indicatea well damped mode. With any algorithm, there will beestimated modes that are numerical artifacts, and not truesystem modes. To discriminate between these two results,a “modal energy” method can be used to determine whichmode in the intearea mode range has the largest energy inthe signal [23].

In [25] spectral processing techniques have have beensuccessfully applied to characterize the small signal oscil-latory modes in the US WECC interconnection, the USEastern Interconnection, the Nordic Power System, andFDR data from the Nigerian power system. Analysis of

data from the US WECC and the Nordic Power Systemrevealed an interesting feature present in the data [25].Fig. 10 shows a Welch Spectrogram bearing a very narrowband signals appearing in the data about 0.28 Hz, whichmay be interpreted as a forced oscillation. We refer thereaders to [25] for these analyses. Here we further de-scribe possible issues for mode meter algorithm issues dueto the presence of these oscillations.

A more careful inspection of the Welch spectrogramreveals that presumably another sinusoid is present atabout 0.18 Hz for the time frame of 0-3 hrs. A this point,it should be realized that it is very likely that both of thesecomponents are harmonics of a fundamental sinusoid of0.09 Hz. In fact, in Fig. 10, it is possible to observetraces of other harmonics at n = 2-6, 8-13, 15-19 and 21,where n is the number of the harmonic. The correspond-ing harmonic frequencies are fn=0.18, 0.28, 0.37, 0.46,0.55, 0.74, 0.83, 0.92, 1.02, 1.11, 1.20, 1.39, 1.48, 1.57,1.66, 1.76 and 1.94 Hz¶. The origin of the sinusoid and itsharmonics is unknown; it could be possibly due to a pro-cess in the system (as a control system going into a limitcycle), aliasing from higher-frequencies, or communica-tion or measurements issues. It is important to note thatsome of the sinusoid harmonics are superimposed over the“true” system modes.

0 8 16 24 32 40 48−10

0

10

20

Damping (%)

Sinusoid 2Mode Meter Estimates from RAD132 P

4 (0.275−0.277Hz Range)

0 8 16 24 32 40 480.25

0.26

0.27

0.28

0.29

0.3

Frequency (Hz)

0.4

Figure 11: Damping and frequency estimation of the 0.18 and 0.27 Hzcomponents.The time is given in hours in (UTC) starting from 00:00:00hrs, local time is given in UTC+1 hr on 03-20-2008 and 03-21-2008

For the component 0.28 Hz in Fig. 11 the average fre-quency computed over 48 hrs. is 0.276 Hz (with a smalldeviation of 0.004 Hz), and an average damping of 0.59%(with standard deviation of 1.08%). For this componentthe estimation results give confidence (a much lower stan-dard deviation) that a forced sinusoid does exist, and fur-thermore, that the estimated damping of these componentsis zero.

The main concern is not necessarily that these forcedsinusoids exist, but more importantly is to realize thatsome of the sinusoid harmonics are superimposed over the“true” system modes. Hence, it should be expected thatthe damping estimates for “true” system modes will be af-fected by the presence of forced oscillations. Similar to the

¶Note that frequencies 0.92, 1.02, 1.39, 1.76 and 1.94 Hz are difficult to observe with the dB scale used for the temperature map in Fig. 10)

results shown in Fig. 11, the average frequency and damp-ing for each of the modes discussed in the last section werecomputed. Table 2 shows the computed averages alongwith their respective standard deviations. While it is pos-sible to have confidence on the estimates for the frequency,it is not possible to do so for the damping estimates for the“true” system modes (Mode 1 through Mode 3). Thus, itis possible to conclude that the presence of the superim-posed harmonic components of the forced oscillation re-duces the damping estimates for each of these frequencies.As mentioned above, the mode meter algorithm is not ca-pable of resolving the portion of the frequency spectrumthat is due to ambient load variation from the portion dueto the forced oscillations, and as a consequence the analystshould scrutinize carefully of the resulting damping esti-mates. The difficulties exposed above pose a new researchchallenge to improve the accuracy of mode damping es-timates in mode meter algorithms and further research isneeded to address these issues for the deployment of modemeter tools in control centers.

Mode Signal f (Hz) σf d(%) σd

1 P1 0.361 0.004 4.38 7.39P3 0.360 0.006 8.06 9.11P4 0.360 0.004 5.60 7.77

2 P1 0.555 0.006 4.92 9.19P3 0.555 0.007 6.43 8.79P4 0.555 0.006 5.17 6.44

3 P1 0.829 0.004 2.53 5.43P3 0.830 0.005 3.86 6.86P4 0.829 0.004 3.29 5.50

Table 2: Statistics of Mode Meter Estimates for 48 hrs.

5 Conclusions

In this paper we have presented synchronized pha-sor measurement applications and phasor data concentra-tor approaches. We have highlighted phasor measurementdata issues and how applications can be designed to takethem into account. These results can serve to derive re-quirements needed for the actual implementation of theapplications in a control center. In addition, we havealso outlined a phasor data concentrator approach that canmanage and exploit phasor measurement data and serveas a phasor gateway. Thus, our overall aim with this pa-per has been to highlight that the successful transition oftoday’s bulk power delivery system into “Smart Trans-mission Grids” will largely depend on the developmentand implementation of synchronized phasor measurementdata-based applications, and on the continuous improve-ment of the information and communication infrastruc-tures supporting them.

6 Acknowledgement

Joe H. Chow is supported in part by a grant from theGlobal Climate and Energy Project at Stanford University,and in part by the Power System Research Consortium ofFirstEnergy, ISO-NE, NYISO, and PJM.

Luigi Vanfretti is supported by the STandUP for En-ergy collaboration initiative, and the KTH School of Elec-trical Engineering.

REFERENCES[1] IEC 60870-6 / TASE.2: ICCP. IEC Standard 60870-6-802, 2005.

[2] D. Trudnowski and J.W. Pierre. In A. R. Messina, editor, Inter-areaOscillations in Power Systems: A Nonlinear and NonstationaryPerspective, Power Electronics and Power Systems, chapter SignalProcessing Methods for Estimating Small-Signal Dynamic Proper-ties from Measured Responses, pages 1–36. Springer, 2009.

[3] A. Armenia and J. Chow. A flexible phasor data concentrator de-sign leveraging existing software technologies. IEEE Transactionson Smart Grid,, 1(1):73 –81, jun. 2010.

[4] D. Atanackovic, J. Clapauch, G. Dwernychuk, J. Gurney, andH. Lee. First steps to wide area control. IEEE Power and EnergyMagazine, 6(1):61 –68, 2008.

[5] D. Bakken, C. Hauser, and H. Gjermundrod. Delivery Require-ments and Implementation Guidelines for the NASPInet Data Bus.IEEE SmartGridComm, IEEE International Conference on SmartGrid Communications, oct. 2010.

[6] A. Bose. Smart transmission grid applications and their supportinginfrastructure. IEEE Transactions on Smart Grid, 1(1):11 –19, jun.2010.

[7] T. Bradley, C. Brown, and A. Malis. RFC 1490: MultiprotocolInterconnect over Frame Relay, 1993.

[8] M. Chenine, L. Nordstrom, and P. Johnson. Factors in assessingperformance of wide area communication networks for distributedcontrol of power systems. IEEE Lausanne Power Tech, 2007, pages1682 –1687, jul. 2007.

[9] J. De La Ree, V. Centeno, J. Thorp, and A. Phadke. Synchronizedphasor measurement applications in power systems. IEEE Trans-actions on Smart Grid, 1(1):20 –27, jun. 2010.

[10] A. Ghassemian and B. Fardanesh. Phasor assisted state estimationfor nys transmission system. In IEEE PES Power Systems Confer-ence and Exposition,, pages 1 –8, 2009.

[11] H. Gjermundrod, D. Bakken, C. Hauser, and A. Bose. GridStat:A Flexible QoS-Managed Data Dissemination Framework for thePower Grid. IEEE Transactions on Power Delivery, 24(1):136 –143, 2009.

[12] A. Gomez-Exposito and A. de la Villa Jaen. Two-level state es-timation with local measurement pre-processing. Power Systems,IEEE Transactions on, 24(2):676 –684, May 2009.

[13] A. Hauer, D. Trudnowski, and J. DeSteese. A Perspective onWAMS Analysis Tools for Tracking of Oscillatory Dynamics.IEEE Power Engineering Society General Meeting, 2007., pages1–10, June 2007.

[14] Y. Hu. Data bus technical specifications for NASPInet, 2008.

[15] W. Jiang, V. Vittal, and G. Heydt. A distributed state estimator uti-lizing synchronized phasor measurements. IEEE Transactions onPower Systems, 22(2):563 –571, May 2007.

[16] K. Martin and J. Carroll. Phasing in the technology. Power andEnergy Magazine, IEEE, 6(5):24 –33, 2008.

[17] A. Monticelli and A. Garcia. Reliable bad data processing for real-time state estimation. IEEE Transactions on Power Apparatus andSystems, PAS-102(5):1126 –1139, May 1983.

[18] M. Parashar and J. Mo. Real Time Dynamics Monitoring Sys-tem (RTDMS): Phasor Applications for the Control Room. In42nd Hawaii International Conference on System Sciences, 2009(HICSS ’09), pages 1–11, Jan. 2009.

[19] A. Phadke and R. de Moraes. The wide world of wide-area mea-surement. IEEE Power and Energy Magazine, 6(5):52 –65, sep.2008.

[20] A. G. Phadke, J. S. Thorp, and K. J. Karimi. State estimationwith phasor measurements. IEEE Transactions on Power Systems,1(1):233 –238, 1986.

[21] E. Rosen, A. Viswanathan, and R. Callon. RFC 3031: Multiproto-col Label Switching Architecture, 2001.

[22] J. Thorp, A. Phadke, and K. Karimi. Real time voltage-phasor mea-surement for static state estimation. IEEE Transactions on PowerApparatus and Systems, PAS-104(11):3098 –3106, 1985.

[23] D. Trudnowski, J. Pierre, N. Zhou, J. Hauer, and M. Parashar. Per-formance of Three Mode-Meter Block-Processing Algorithms forAutomated Dynamic Stability Assessment. IEEE Transactions onPower Systems, 23(2):680–690, May 2008.

[24] L. Vanfretti, J. H. Chow, S. Sarawgi, and B. B. Fardanesh. Aphasor-data-based state estimator incorporating phase bias correc-tion. IEEE Transactions on Power Systems, 2010. Accepted forfuture publication.

[25] L. Vanfretti, L. Dosiek, J. W. Pierre, D. Trudnowski, J. H. Chow,R. Garcıa-Valle, and U. Aliyu. Application of ambient analy-

sis techniques for the estimation of electromechanical oscillationsfrom measured PMU data in four different power systems. Euro-pean Transactions on Electrical Power, 2010. Accepted for futurepublication.

[26] T. Yang, H. Sun, and A. Bose. Transition to a two-level linearstate estimator;part i: Architecture. IEEE Transactions on PowerSystems, 2010.

[27] L. Zhao and A. Abur. Multi area state estimation using synchro-nized phasor measurements. IEEE Transactions on Power Systems,20(2):611 – 617, May 2005.

[28] J. Zuo, R. Carroll, P. Trachian, J. Dong, S. Affare, B. Rogers,L. Beard, and Y. Liu. Development of TVA SuperPDC: Phasorapplications, tools, and event replay. In Power and Energy SocietyGeneral Meeting - Conversion and Delivery of Electrical Energyin the 21st Century, 2008 IEEE, pages 1–8, July 2008.